Many systems are required to comply with safety standards. For example, automotive electric and electronic systems are required to comply with the ISO 26262 standard.
Usually analysis of a system is done in the context of a specific application being considered and, in the automotive sector, examples of applications include adaptive cruise control, airbag and braking.
Application analysis is conducted to understand a given application and to identify system components, how the components interact and how each component is used. Design partitioning is used to break down each hardware component into parts. If necessary, parts can be broken down further into sub-parts to aid the analysis. Fault probabilistic characterisation is carried out to identify plausible faults for each part or sub-part. Fault impact analysis is then performed to analyse the impact of each fault on the application. Safety mechanisms analysis is then conducted to determine how safety mechanism(s) cover significant faults. Finally, fault coverage analysis is carried out to determine the coverage offered by a given safety mechanism. This results in a set of parameters for each system for the application.
The steps depend on the application under consideration. Thus, if the application changes, then the design may need to be re-partitioned considering, for example, a different level of sub-parts and a new set of plausible faults are identified. The impact of each fault, analysis of safety mechanisms and the coverage of safety mechanisms may also need to be re-assessed.
Microcontroller units (MCUs) and application specific integrated circuits (ASICs) are widely used in automotive-related applications.
EP 1980964 A1 describes a method of performing failure mode and effects analysis (FMEA) on integrated circuits. The method includes preparing a FMEA database which includes FMEA statistics and computing FMEA results using the FMEA database. However, the FMEA statistics are specific to a particular usage. Thus, if the usage changes, then a new set of FMEA statistics is prepared. This tends to be inefficient for MCUs.